- AI For All
- Posts
- Llama 3.1 Dethrones GPT-4
In partnership with
Hello readers,
Welcome to another edition of This Week in the Future! Meta released Llama 3.1, with the 405B version dethroning GPT-4, and OpenAI unveiled SearchGPT, an AI answer engine that is their answer to Perplexity and Google Search.
Llama 3.1
Image made with DALL·E 3
Meta has released Llama 3.1 with 8B, 70B, and 405B parameter variants. The 405B version surpasses GPT-4, a significant achievement for open source, only to then be immediately surpassed by Mistral Large 2 (also open source) on some benchmarks. Here’s how Llama 3.1 405B stacks up against the leading closed source models:
Mark Zuckerberg followed up the release of Llama 3.1 with a statement on open source AI where he explains his long-term vision for the AI ecosystem and why Meta has taken the route that it has. He even shows that he has a sense of humor when he writes, “Meta’s business model is about building the best experiences and services for people.”
Meanwhile, OpenAI is truly without a moat and may lose $5 billion this year. Accenture will be offering enterprises customized versions of Llama 3.1 with NVIDIA AI Foundry. With Llama 3.1 and Mistral Large 2, David is beating Goliath.
On the subject of OpenAI…
SearchGPT
OpenAI has unveiled a prototype of SearchGPT, their answer to Perplexity and yet another move against Google. In their announcement, they smartly emphasize prominently citing and linking to publishers in searches. There is currently a waitlist to test SearchGPT. The best features will eventually be weaved into ChatGPT.
If AI search is the future, it will need to be done in a way that incentivizes publishers to keep creating original content. Otherwise, AI will increasingly only have AI-generated content to train on, leading to catastrophic model collapse.
🔥 Rapid Fire
OpenAI improves model safety with Rule-Based Rewards
Google DeepMind’s AI wins silver medal in Math Olympiad
Google upgrades Gemini with 1.5 Flash for faster responses
Bing announces generative search to compete with Google
Cohere introduces Rerank 3 Nimble for enterprise RAG
Stability AI introduces Stable Video 4D, a video-to-3D model
Generative AI found to help individual creativity but harm diversity
Study reveals rapid increase in web domains blocking AI models
Coatue explains why robotics won’t have a ChatGPT moment
Google faces scrutiny on AI from analysts in earnings call
Shares in tech companies drop in U.S. and Asia as AI stocks slide
Elon Musk announces world’s ‘most powerful’ AI training cluster
Microsoft and Mass General Brigham advance AI medical imaging
US lawmakers send letter to OpenAI demanding answers
FTC, DOJ, and others issue statement on competition in AI
FTC investigates companies using AI for surveillance pricing
Ukraine rushes to create AI-enabled war drones
📖 What We’re Reading
“While there’s been a lot of chatter about the benefits of generative AI (Gen AI), many organizations are not quite sure about how to implement it to boost productivity and effectively integrate it into their employee’s workflows. A recent SAS study highlighted this uncertainty: while nearly 20 percent of business leaders surveyed (in the U.S. and Canada) had fully implemented Gen AI within their organizations, nearly nine out of 10 senior technology decision-makers didn’t understand how Gen AI impacted business processes.”
💻️ AI Tools and Platforms
Salt AI → Visual, collaborative AI development
Weights & Biases → AI developer platform
Veritone → Enterprise AI software provider
Triple Whale → AI data platform for ecommerce
hoop.dev → AI data masking for databases
We put your money to work
Betterment’s financial experts and automated investing technology are working behind the scenes to make your money hustle while you do whatever you want.
What did you think of today's newsletter?We value your feedback, so please take time to let us know what you liked and what could be improved. |